Network management platform based on mobile agents

Reviewer: Damianos Gavalas

This paper addresses an interesting topic, which received particular attention in the late 1990s. Research on this subject has declined during the past few years, however, due to security concerns arising from the mobile agent (MA) paradigm, and to the lack of killer applications in the field of network, systems, and service management. A management platform involving both mobile and fixed agents is presented. Four classes of agents are defined, which mainly perform monitoring tasks. The prototype implementation is built on a general-purpose agent platform, Toshiba's Bee-gent agent platform, not optimized for management applications. Experimental results reveal that traditional centralized management (using the HP OpenView tool) outperforms the presented architecture for small-to-medium sized networks. However, the latter exhibits superior scalability, distributes systems resources overhead, and prevents network bottlenecks around the manager station. The paper does not introduce any particularly novel ideas in the area MA-based network management. Its most interesting aspect is the proposal of ontology-driven agents, whereby agents are not preprogrammed, but instead acquire new knowledge and intelligence by accessing an ontology database. This concept is not detailed in the paper, however, and concrete ontology examples are not given. Online Computing Reviews Service

Reviewer: Naga R Narayanaswamy

This paper discusses a network management framework that is different from the traditional simple network management protocol (SNMP) model. The new framework seeks to remove the shortcomings of the current model, for example large communication overheads and over-centralized management structures. The authors propose a management platform based on agent technology, involving fixed intelligent agents coordinating a number of mobile agents to achieve a scalable flexible management infrastructure. The paper explains the platform, and presents a description of its evaluation and an analysis of the results. It would have been better if the authors had divided the design and implementation aspects more thoroughly; they are intermixed at times. Though the concepts are explained well, at times, the paper requires a second pass when reading since terms are used before they are defined. The idea of the paper is well taken, but, to achieve the authors’ objective of wide acceptance, it will have to go through the Internet Engineering Task Force (IETF) forums to get a standardization approval. No reference to this is presented in the paper. Because SNMP is so much a part of the current management structure, any success at augmenting it will need the ratification of the IETF committee. Online Computing Reviews Service